docker_container.running fails with an address-already-in-use error when a
container holds a stale sobridge endpoint - either after its IP changes in
pillar or after an unclean shutdown leaves the endpoint behind.
Add the clear_stale_endpoint macro and call it from every enabled.sls whose
container attaches to sobridge with a static IP. The macro force-disconnects
the container from the network only when its current endpoint IP does not
match the one salt is about to assign, so it is a no-op on a normal highstate.
The strelka states use state IDs that differ from their container names, so
those calls pass state_id explicitly. nginx resolves its IP through
container_config, which varies by role.
Josh Patterson
cd8f7f99e3